A CMOS Fully-Integrated Wireless Power Receiver for Autonomous

A CMOS Fully-Integrated Wireless Power Receiver for
Autonomous Implanted Devices
Fabian L. Cabrera, and F. Rangel de Sousa
Electrical Engineering Department
Federal University of Santa Catarina
Florian´opolis-SC, 88040-900, Brazil.
E-mail: [email protected], [email protected]
I. I NTRODUCTION
There is a growing interest in the use of electronic implanted
devices for applications such as sensor networks, medical systems
and remote instrumentation. These applications have in common the
demand for miniaturized and autonomous devices, with difficult or
even no access for energy source replacement. In this context, the
batteries are unsuitable and the energy must be supplied by means
of wireless power transferring (WPT) [1]. Two magnetically coupled
inductors are usually employed for wireless power transferring, one
of them is packaged together with the embedded electronic circuit.
The on-chip integration of the receiving inductor must be pursued in
order to reduce size and cost, to increase mechanical robustness and
to achieve mass production.
While inductive links implemented with printed boards have been
widely studied [2−4], the integration of the WPT system in a single
silicon die is rare. In [2], the authors explore the design of inductive
links for powering implants in several scenarios, including the effect
of biological tissue and considering the possibility of using an
integrated inductor in a CMOS process, but they did not implement
a full receiver. The implementation of a complete WPT system,
operating at 1 GHz, was reported in [5], nevertheless the inductors
are out of the chip. The efficiency of the 3-stage rectifier was 65%,
however, the losses on the matching network were not taken into
account. A rectifier presenting 86% of efficiency was brought in [6].
Nonetheless, they designed a single-stage circuit, that is difficult to
match with simple networks and requires a high-voltage amplitude
signal at the input. The theoretical work discussed in [7] suggest that
the optimal frequency for powering implants is in the GHz scale,
without considering that the frequency where occurs the maximum
quality factor of an inductor depends on its size and on the number of
turns. In [8], a matching network was proposed using an additional
inductor which would not be a good option in an integrated system
because of the poor quality factor of on-chip inductors.
The main objective of this work is to demonstrate the possibility
of a wireless power receiver fully integrated, constrained to an area
of 1.5 mm × 1.5 mm in the IBM 0.18 µm CMOS process. The
energy is received on-chip via an inductive link. To achieve this goal,
Zin
Rectifier
Inductive
Link
Matching
Network
ZX
M
I1
RL
I2
L1
L2
R1
R2
vtest
RX + jXX
Abstract—In this paper we present the design of a wireless power
receiver fully integrated. The circuit was constrained to occupy a silicon
area of 1.5 mm × 1.5 mm in a 0.18 µm RF-CMOS process. The main
target was to optimize the part of the power transfer efficiency concerning
only the receiver side. In that way, we optimized the quality factor of the
integrated inductor, the impedance matching conditions and the rectifier
efficiency. The simulated quality factor of the integrated inductor was
22 using a link frequency of 1 GHz. Post-layout simulations of the entire
system shows that the combined efficiency of the impedance matching
network and the rectifier is 57% when the available power at the inductor
is 1 dBm. Moreover, the system uses backscattering to respond to the
transmitter, allowing to infer the total power transfer efficiency.
(a)
(b)
Fig. 1. (a) WPT system. (b) Inductive link model.
an inductor occupying the outermost area of the chip was integrated.
The chip also includes the impedance matching, the voltage rectifier
and a circuit that functions as variable load for the receiver and at
the same time generates a response, which will be communicated to
the transmitter using the same inductive link.
II. W IRELESS P OWER T RANSFER E FFICIENCY
The WPT system is responsible to receive and process the energy
delivered to the implanted device. As shown in Fig. 1(a), it is
composed of an inductive link, a matching network and a rectifier.
The link is formed by two inductors as shown in Fig.1(b). Each
inductor L1(2)
√ has a series equivalent resistance R1(2) , modeling the
losses. M =k L1 L2 is the mutual inductance and k is the magnetic
coupling factor with values ranging from 0 to 1. The equivalent
load impedance is ZX =RX +jXX . The link power efficiency (η0 )
is defined as the ratio between the power at the load (|I2 |2 RX ) and
the power delivered to the link (|I1 |2 <e{Zin }), where I1 and I2 are
the mesh currents and <e{Zin } is the real part of Zin , the link input
impedance. Solving I1 and I2 by mesh analysis we obtain:
η0 =
(ωM )2 RX
.
2
R1 [(ωL2 +XX ) +(R2 +RX )2 ]+(ωM )2 (R2 +RX )
(1)
This function is maximized with respect to XX , when XX =−ωL2 .
Under this condition, we can calculate the reciprocal of the link
efficiency (1/η) as:
1
1 R2
RX
R2
R1 R2
=
=
+
2
+
+
+ 1.
η
η0 (ωM )2 RX
R2
RX
XX =−ωL2
(2)
Recognizing that ωL1(2) /R1(2) is Q1(2) , the quality factor of the
primary (secondary) inductor, defining p=R2 /RX and using (2), we
can write the reciprocal of efficiency (ηT ) of the WPT system as:
1
1
1 1 1
1
=
(p + 2 + ) + p + 1 ,
(3)
ηT
ηRT k2 Q1 Q2
p
where ηRT was included to account for the rectifier efficiency. The
matching network was considered lossless in (3). In this work, we
are only concerned with the implanted part of the WPT system, then,
the product k12 Q11 is considered as an input parameter. Moreover, we
assume that the link is weakly coupled, since it is the worst case
RF+
VDD
Impedance
Matching
Inductor
RF-
Rectifier
as simple as possible to reduce the sensitivity to process variability.
After these assumptions, we decided for using a single matching
capacitor (CM ), connected in parallel to the inductor.
Variable
Load
sw
Response
Signal
Vs
The system designed in this section is illustrated in Fig. 2. It
is composed of a WPT receiver and a variable load. The energy
is received by an integrated inductor occupying the outermost area
of the chip. Following the inductor, a passive network provides
conjugate matching between the link and rectifier impedance. Then,
a rectifier converts the RF input signal to a DC voltage that powers
a ring oscillator, which performs as a variable load. The oscillator is
used to generate a signal that is sent back to the transmitter by the
inductive link. The oscillator frequency carries the information about
the power received by the WPT system.
A. Integrated Inductor
The integrated inductor was implemented in the uppermost metal
level of an 180 nm RF-CMOS technology. The inductor design
variables are the turns number nind2 , the line width wind2 and the
turns separation sind2 , as shown in Fig. 3(a). The external diameter
(dext2 ) must be as large as possible to maximize the magnetic flux
enclosed by the inductor, in this case that value is 1460 µm.
22
sind2 nind2
21
−−
19
30µm 2
max(Q2)
10µm 2
10µm 3
18
sind2
1
20
30µm 3
17
16
dext2
100
150
200
250
300
350
400
wt2 [µm]
(a)
(b)
Fig. 3. (a) Two-turns inductor. (b) Maximum Q2 for several inductors.
In order to find the inductor dimensions that maximize its quality
factor, we did electromagnetic simulations using the software EMPRO from Agilentr , where the variables wt2 and sind2 were swept
for nind2 =1,2 and 3. The maximum quality factor for each inductor
is plotted in Fig. 3(b) as function of wt2 . According to the figure the
inductor with nind2 =1 and wind2 =250 µm has the highest quality
factor (Q2 =22.4), which happens when the link frequency is 1.04
GHz. As a consequence, this was the frequency chosen for operating
the system.
B. Impedance matching network
Two factors were considered for choosing the matching network
topology: (i) The network should not include inductors, because
integrated inductors have poor quality factor. (ii) The network must be
C2
Inductive
Link
III. S YSTEM D ESIGN
wind2
ART
CM
L2
RL
RRT
of efficiency and leads to a design independent from the primary
inductor properties. So the optimization of the total efficiency is
restricted to optimizing the values of Q2 , p and ηRT .
wt2
RL’
R2
Fig. 2. Diagram of the integrated system.
davg2
ZX
L2
CRT
Matching
Capacitor
Rectifier
ZX
R2
CX
Vs
RX
Load
(a)
(b)
Fig. 4. Model for the power receiver: (a) Extended. (b) Simplified.
An electrical model to calculate the impedance matching is presented in Fig. 4(a), where C2 models the self-resonance frequency
of the inductor, CRT is the input capacitance of the rectifier, RRT
represents the rectifier losses, ART is the voltage gain of the rectifier,
and RL is the load. The model of Fig. 4(a) can be simplified at
the frequency of interest resulting in Fig. 4(b). The conditions for
impedance matching are met by choosing the values of CM and RL
so that CX =1/(ω 2 L2 ) and RX =R2 , for a given rectifier.
Using RL to match the receiver means that there is a received
power level that optimizes efficiency. A problem arises when the
maximum efficiency occurs at a high level of power. There are two
ways to move the optimal power to a lower value: (i) Increase ART by
increasing the number of stages in the rectifier, because the equivalent
0
will decrease and the matching condition will occurs
resistance RL
at a lower power level; (ii) increase the equivalent resistance of the
inductor by reducing wind2 or by increasing nind2 , decreasing Q2
and as a consequence, decreasing the total efficiency in both cases.
Since the main objective in this design is to maximize the efficiency,
we preferred to use an inductor with the best quality factor and
keep the optimal power level at a moderated value (1 dBm) by
implementing a four-stage rectifier.
The matching efficiency can be evaluated by inspecting the ratio
between the input power at the rectifier PRT and the available power
from the inductor Pavs , defined as:
PRT
ηM =
.
(4)
Pavs
C. Rectifier
The design of the rectifier considered the efficiency as the main
figure of merit. The rectifier topology was based in [9] due to its
high efficiency and simplicity and it is is shown in Fig. 5(a). The
number of stages was chosen to be four, in order to have 2.5 V at
the output when the available power is 1 dBm. The block diagram
of the four-stages rectifier is shown in Fig. 5(b).
in1
RF+
in2
RFV-
V+
RT1
RT2
RT3
gnd
(a)
RT4
CL
VDD
RL
(b)
Fig. 5. Rectifier: (a) CMOS schematic of each stage. (b) Block diagram.
All transistors in Fig. 5(a) have the minimal length (0.18 µm)
and width WM =30 µm, which corresponds to the best compromise
between ηM and ηRT . The NMOS transistors are triple-well devices,
which permits to connect their sources to their body terminals. The
rectifier efficiency was simulated as function of the available power
from the inductive link (Pavs ) for several values of RL and the results
are shown in Fig. 6(a). We observe in the figure that for each value
of RL there is a value of Pavs that maximizes the efficiency of the
rectifier ηRT . We extracted the point of maximum of each curve in
Fig. 6(a) and plotted them against Pavs into Fig. 6(b). The values of
ηM and ηM ηRT corresponding to the points extracted were plotted
in Fig. 6(c) and in Fig. 6(d) respectively. Although the rectifier alone
can have efficiencies greater than 60% for a wide rank of Pavs values,
the highest ηM .ηRT values are observed for Pavs between 0 and 2
dBm and RL between 7 kΩ and 10 kΩ. Outside of this band, total
efficiency decreases due to impedance mismatch between the link and
the rectifier. Post-layout simulations showed that the best impedance
matching is obtained when Pavs =1 dBm and the load is RL =10 kΩ,
resulting in VDD =2.5 V .
in the oscillator is shown in Fig. 7(b). In addition to the NMOS
and PMOS transistors forming a conventional inverter, we added
transistors with sources connected to the drains acting as capacitors
to decrease the frequency of oscillation. We also added an 1 M Ω
resistance in parallel with the PMOS transistor to allow the operation
of the circuit at low VDD values. So the designed circuit can operate
for VDD between 0.7 V and 3.6 V , as verified by simulations with
typical-case models.
VDD
VDD
D
Level
Conv. sw
lv
Q
in
1MΩ
out
4
1.2
gnd
7-stage
ring oscillator
8
12
2
1.2
sw
8
12
gnd
(a)
(b)
Fig. 7. Variable load: (a) Blocks diagram. (b) Inverter.
60
70
50
60
4kΩ
7kΩ
10kΩ
20kΩ
40kΩ
80kΩ
160kΩ
40
30
20
10
−10
−5
0
Pavs [dBm]
max(ηRT) [%]
ηRT [%]
(40kΩ)
(20kΩ) (10kΩ)
(4kΩ)
50
(160kΩ)
40
30
5
−10
−5
0
Pavs [dBm]
(7kΩ)
(20kΩ)
(4kΩ)
50
ηRT.ηM [%]
ηM [%]
20
(10kΩ)
60
(4kΩ)
40
(40kΩ)
60
40
(7kΩ)
(20kΩ)
80
5
(b)
(10kΩ)
(40kΩ)
30
(80kΩ)
(80kΩ)
20
(160kΩ)
−10
(160kΩ)
10
−5
0
Pavs [dBm]
5
The frequency of the response signal (fsw ) can not be so high,
because it will be band-pass filtered by the transmitter resonant tank.
Simulations at system level done with ADS shows that fsw must be
lower than 10 MHz. On the other hand, low values of fsw would
require a large capacitor CL according to (5):
VDD
,
(5)
4.RL .fsw .∆V
where ∆V is the variation of VDD due to charging and discharging
CL at fsw frequency. We chose ∆V =0.1 V and fsw =3.5 MHz at the
nominal value of VDD (2.5 V ), which leads to a value of CL =180
pF .
When the switch at the terminals of the inductor is open (sw=0), by
energy conservation, half of the current supplied by the rectifier goes
to the variable load (RV ) and the other half charges the capacitor CL ,
so the equivalent load RL is RV /2. The dimensions for transistors
in Fig. 7b were chosen to adjust RV =20 kΩ and fsw =3.5 MHz for
VDD =2.5 V .
The switch used for load modulation is implemented as shown
in Fig. 8. Two NMOS transistors (one for 1.8 V signal and the
other for 3.3 V signal) are connected between the terminals of the
inductor. Their lengths were kept at the respective minimum values
for decreasing the ON resistance, while their widths were designed
for keeping the ON-OFF amplitudes ratio (AON /AOF F ) higher than
10. Other transistors were used to tie down to ground the RF+ and
RF-, aiming at decreasing the ON resistance of the main switches.
CL =
20
(a)
100
(7kΩ)
(80kΩ)
−10
(c)
−5
0
Pavs [dBm]
5
(d)
Fig. 6. (a) Rectifier efficiency for different RL values. (b) Maximum ηRT .
(c) Efficiency in power transferring due to impedance mismatch. (d) Matching
and rectifier efficiency.
RF+
D. Variable Load and Response Generation
In order to measure the efficiency of the system, we designed a
non-intrusive test strategy. Instead of a fixed resistance, we chose
an oscillator as the load because it provides a value of RL that
is a function of the input power, while the output of the oscillator
can activate a backscattering device. Thus, we can conclude that the
oscillator is a variable load, dependent on the input power. In addition,
by measuring the frequency of the envelop of the backscattered signal,
we can estimate the power that is being received with no requirement
of wires.
The variable load is based on a seven-stage ring oscillator whose
frequency and current consumption vary with the supply voltage,
as shown in Fig. 7(a). Following the ring oscillator, a flip-flop is
employed to divide by two the oscillator frequency to ensure a 50%
duty cycle for the sw signal. The schematic of the inverter used
RF-
sw
swlv
sw
1
0.4
64
0.4
1
0.4
sw
swlv
25
1 0.18 1
0.18 0.18
swlv
gnd
Fig. 8. Switch for load modulation.
IV. R ESULTS
The layout of the system can be seen in Fig. 9. It occupies a silicon
surface 1.5 mm × 1.5 mm. In addition to the receiving inductor and
to the full system, an extra cell containing only the variable load was
included for characterization purposes.
TABLE I
C ORNER SIMULATIONS OF THE INTEGRATED SYSTEM .
Pavs
VDD
[dBm]
[V]
Typical case
-7
1.1
1
2.5
6
3.5
Fast case
-7
1.2
1
2.4
6
3.3
Slow case
-7
1.0
1
2.6
6
3.6
fsw
AON
AOF F
ηM .ηRT
[MHz]
[V]
[mV]
[%]
0.6
3.5
5.2
0.5
0.9
1.3
29
33
72
17
57
54
1.1
3.9
5.7
0.5
0.9
1.2
22
34
60
28
59
54
0.3
3.2
4.9
0.5
0.9
1.3
42
40
71
10
55
54
Fig. 9. Layout of the chip.
1
0
2
1
0
sw [V]
implemented in a constrained silicon area of 1.5 mm × 1.5 mm,
even including the receiving inductor. By means of a backscattering
device, the system responded to the level of the received power,
allowing for the estimation of its total efficiency. In addition, a
design methodology was applied in order to establish the trade-offs
between the design variables, the resource restrictions and the highest
efficiency. We found that the best inductor should have a single turn
and its highest quality factor was reached around 1 GHz, keeping the
possibility of powering implants at UHF frequencies. The combined
efficiency of the matching network and the rectifier was 57%.
ACKNOWLEDGMENT
This work was partially supported by CNPq, INCT NAMITEC,
CAPES and MOSIS.
R EFERENCES
0.5
VDD [V]
RF+(−) [V]
The complete system was simulated after layout parasitic extraction
and the transient response is shown in Fig. 10. In this simulation, the
power available at the inductor is 1 dBm and typical-case models
for MOS transistors were used. The oscillator start-up is facilitated
by the characteristic of its variable power consumption, which is low
for low VDD values. It takes approximately 5 µs for VDD to reach
the steady-state at 2.5 V . In steady state, VDD suffers variations as
expected due to the charging-discharging process of capacitor CL at
3.5 MHz. We can see that the signal at the terminals of the inductor
(RF+ e RF-) is correctly attenuated when sw is high, implying a
correct backscattering.
2
1
0
1
2
3
4
5
6
t [µ s]
Fig. 10. Transient simulation of the receiver.
In order to verify the performance of the circuit, we did corner
simulations and the results are summarized in Table I. The efficiency
in the last column was calculated by measuring fsw and associating
it to the power consumed by RV . That power value was multiplied
by 2, because the rectifier only receives power half of the time. From
the table we can conclude that the system can receive and respond
correctly for Pavs between -7 dBm and 6 dBm.
V. C ONCLUSION
In this paper we presented an integrated circuit designed to
corroborate the feasibility of a wireless power transfer receiver, fully
[1] A. Yakovlev, Sanghoek Kim, and A. Poon, “Implantable biomedical
devices: Wireless powering and communication,” Communications Magazine, IEEE, vol. 50, no. 4, pp. 152–159, 2012.
[2] M. Zargham and P.G. Gulak, “Maximum Achievable Efficiency in NearField Coupled Power-Transfer Systems,” Biomedical Circuits and Systems, IEEE Transactions on, vol. 6, no. 3, pp. 228–245, 2012.
[3] R.R. Harrison, “Designing Efficient Inductive Power Links for Implantable Devices,” in Circuits and Systems, 2007. ISCAS 2007. IEEE
International Symposium on, 2007, pp. 2080–2083.
[4] Uei-Ming Jow and M. Ghovanloo, “Design and Optimization of Printed
Spiral Coils for Efficient Transcutaneous Inductive Power Transmission,”
Biomedical Circuits and Systems, IEEE Transactions on, vol. 1, no. 3,
pp. 193–202, 2007.
[5] S. O’Driscoll, A.S.Y. Poon, and T.H. Meng, “A mm-sized implantable
power receiver with adaptive link compensation,” in Solid-State Circuits
Conference - Digest of Technical Papers, 2009. ISSCC 2009. IEEE
International, 2009, pp. 294–295,295a.
[6] M. Zargham and P.G. Gulak, “High-efficiency CMOS rectifier for fully
integrated mW wireless power transfer,” in Circuits and Systems (ISCAS),
2012 IEEE International Symposium on, 2012, pp. 2869–2872.
[7] A.S.Y. Poon, S. O’Driscoll, and T.H. Meng, “Optimal Frequency for
Wireless Power Transmission Into Dispersive Tissue,” Antennas and
Propagation, IEEE Transactions on, vol. 58, no. 5, pp. 1739 –1750, May
2010.
[8] R.-F. Xue, K.-W. Cheng, and M. Je, “High-Efficiency Wireless Power
Transfer for Biomedical Implants by Optimal Resonant Load Transformation,” Circuits and Systems I: Regular Papers, IEEE Transactions on,
vol. 60, no. 4, pp. 867–874, 2013.
[9] S. Mandal and R. Sarpeshkar, “Low-Power CMOS Rectifier Design
for RFID Applications,” Circuits and Systems I: Regular Papers, IEEE
Transactions on, vol. 54, no. 6, pp. 1177–1188, 2007.